Network: Policy Routing Analysis, DHCP Y. Richard Yang http://zoo.cs.yale.edu/classes/cs433/ 12/04/2018

Outline Admin and recap Network control plane Routing Link weights assignment Routing computation Basic routing protocols (distance vector, link state) Global Internet routing Basic architecture Policy routing analysis Address assignment

Admin Assignment five (written assignment) questions Exam 2 reminder: 7:00-8:30 pm Dec. 11 at AKW 200 and AKW 300

Recap: Basic Routing Protocols Safe, Distributed DV protocols: Issue: Counting-to-infinity Basic idea: use local, sufficient conditions to enforce global no loop condition Analytical techniques: use global invariants to gain understanding of distributed protocols Link state protocols Instead of distributed computing, use distributed state synchronization to avoid the churns of distributed computation Simplicity churns before convergence path exploration; counting-to-infinity not enough information when making routing decision (only next-hops and their distance estimates) While distance-vector protocols are appropriate for small internetworks, and require much less configuration and management - IGRP was a significant improvement over RIP, which had a hop count restriction that limited the size of an internetwork. IGRP supports internetworks with up to 255 hops. IGRP: Upon router startup, IGRP broadcast request messages to other routers. Upon receipt of the message, router send their routing tables to the startup router. Routing tables contain entries for each of the networks that a router can reach. At regular intervals, routers automatically transmit their routing tables to other routers which in turn update their own routing tables if there are changes in the topology. A triggered update occurs if the network topology changes. Triggered updates allow quick responses to changes in the network. When routers add entries to routing tables, 1 is added to the hop count to account for the hop between the router and the neighbor from which it received the route information. When a route is entered into the routing table, a timer is set. As routing table updates arrive and a route entry is verified to still be valid, the timer is reset. If a known router fails to appear in an update and in subsequent updates, the timer will run out and the route entry will be purged.

Recap: Global Internet Routing Architecture Inter-AS routers form an overlay Gateway routers participate in intradomain to learn internal routes. Gateway routers of same AS share learned external routes using iBGP. AS C (RIP intra routing) AS B (OSPF intra routing) AS A (OSPF intra routing) b i eBGP a e iBGP Gateway routers of diff. auto. systems exchange routes using eBGP g f d h c

BGP Basic Operations AS1 BGP session AS2 Establish session on TCP port 179 AS1 BGP session Exchange all active routes AS2 while (connection is ALIVE) exchange UPDATE message select best available route if route changes, export to neigh. Exchange incremental updates

BGP Messages Four types of messages OPEN: opens TCP connection to peer and authenticates sender UPDATE: advertises new path vectors (or withdraws old) KEEPALIVE keeps connection alive in absence of UPDATES; also ACKs OPEN request NOTIFICATION: reports errors in previous msg; also used to close connection

Recap: CIDR Address Aggregation to Reduce # Detinations Active BGP Entries (http://bgp.potaroo.net/as1221/bgp-active.html) Internet Growth (http://www.caida.org/research/topology/as_core_network/historical.xml)

Recap: Features of Global Routing Architecture Design Scalability Only a small # of routers (gateways) from each AS in the interdomain level CIDR aggregation reduces amt data to be carried Privacy Interdomain routing carries only path vector, not internal network path Autonomy Autonomous systems have flexibility to choose their own intradomain routing protocols Each network chooses interdomain path according to its own policy

Recap: BGP as a Policy Routing Framework route selection policy: rank egress paths 1 3 0 1 0 select best path routing cache export policy: which paths export to which neighbors controls ingress export path to neighbors Qwest Internet2 AT&T Yale

Outline Admin and recap Network control plane Routing Link weights assignment Routing computation Basic routing protocols (distance vector, link state) Global Internet routing Basic architecture Policy routing analysis

Policy Routing Instability A policy routing system can be considered as a system to aggregate individual preferences, but aggregation may not be always successful. The BAD GADGET example: - 0 is the destination - the route selection policy of each AS is to prefer its counter clock-wise neighbor preferred 2 3 1 2 1 0 2 0 1 3 0 1 0 3 2 0 3 0 4 less preferred Policy (preferences) aggregation fails: routing instability !

General Framework of Preference Aggregation Also called Social Choice Given individual preferences, define a framework (constitution) to aggregate individual preferences: A set of choices: a, b, c, … A set of voters 1, 2, … Each voter has a preference (ranking) of all choices, e.g., voter 1: a > b > c voter 2: a > c > b voter 3: a > c > b A well-specified aggregation rule (protocol) computes an aggregation of ranking, e.g., Society (network): a > c > b https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem

Example: Aggregation of Global Preference Choices (for S->D route): SAD, SBD, SABD, SBAD Voters: S, A, B, D Each voter has a preference, e.g., S: SAD > SBD > SABD > SBAD … A https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem S D B

Global Aggregation Framework Axioms: Transitivity if a > b & b > c, then a > c Unanimity: If all participants prefer a over b (a > b) => a > b Independence of irrelevant alternatives (IIA) Social ranking of a and b depends only on the relative ranking of a and b among all participants Result: Arrow’s Theorem: Any constitution (protocol) that respects transitivity, unanimity and IIA must be a dictatorship. https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem

Proofs of Arrow’s Theorem There are quite a few proofs, and the six-page paper linked on the Schedule page gives three simple proofs. Below, I give the key insight of the proof using approach 1.

The Extremal Lemma Let choice b be chosen arbitrarily. Assume that every voter puts b at the very top or the very bottom of his ranking. Then constitution must as well (even if half voters put b at the top and half at the bottom) Proof: by contradiction. Assume there exist a and c such that constitution has a >= b; b >= c. We can move c above a w/o changing a-b or b-c votes

Step 1: Existence of Pivotal Voter Let choice b be chosen arbitrarily. There exists a voter n* = n(b) who is extremely pivotal for b in the sense that by changing his vote at some profile, he can move b from the very bottom to the very top in the social ranking. Proof: Consider an extreme profile where b is at the bottom of each voter. Consider voter from 1 to n, and we move b from bottom to top one-by-one. The first voter whose change causes b to move to the top is n*

Step 2: n*=n(b) is dictator of any pair ac not involving b Proof Consider a from ac pair. We show that if a >n* c, then society has a > c Let profile before n* moves b to top as profile I Let profile after n* moves b to top as profile II Construct profile III from II by letting n* move a above b; all others can arrange ac as they want, but leave b in extreme position b . Profile I constitution: b bottom b . constitution : b top Profile II 1 2 n* . N constitution : a > b since ab same as I b . a c Profile III b > c since bc same as II

Step 3: n* is dictator for every pair ab Consider c not equal to a or b There exists n(c) who is a dictator of any pair not involving c, such as the pair ab, i.e., For any profile, if a >n(c) > b, a > b for society n(c) must be n* Assume not. Consider Profile I and Profile II. Since n(c) is not n*, n(c) ranking of ab does not change in Profile I and Profile II. When n* changes ab ranking between Profile I and Profile II, the constitution ranking of ab changes. Contradiction.

Outline Admin and recap Network control plane Routing Link weights assignment Routing computation Basic routing protocols (distance vector, link state) Global Internet routing Basic architecture Policy routing analysis Issue: Instability Global aggregation and dictatorship Local aggregation and P-graph

BGP w/ Local Preference BGP preferences are typically local, on egress routing (only on paths starting from itself) Hence the preferences have dependency (priority) The “closer” a node to the destination, the more “powerful” it may be 2 1 3 320 30 210 20 130 10 2 1 0 2 0 1 0 3 0 1 3 0 3 2 0

Complete Dependency: P-Graph 2 1 3 320 30 210 20 130 10 Complete dependency can be captured by a structure called P-graph Nodes in P-graph are feasible paths Edges represent priority (low to high) A directed edge from path N1P1 to P1 intuition: to let N1 choose N1P1, P1 must be chosen and exported to N1 A directed edge from a lower ranked path to a higher ranked path intuition: the higher ranked path should be considered first 2 1 0 2 0 1 0 3 0 1 3 0 3 2 0 Any observation on the P-graph?

P-Graph and BGP Convergence If the P-graph of the networks has no loop, then policy routing converges. intuition: choose the path node from the partial order graph with no out-going edge to non-fixed path nodes, fix the path node, eliminate all no longer feasible; continue Example: suppose we swap the order of 30 and 320 2 1 3 30 320 210 20 130 10 2 1 0 2 0 1 0 3 0 1 3 0 3 2 0

Outline Admin and recap Network control plane Routing Link weights assignment Routing computation Basic routing protocols (distance vector, link state) Global Internet routing Basic architecture Policy routing analysis Issue: Instability Global aggregation and dictatorship Local aggregation and P-graph Economics and interdomain routing patterns

Internet Economy: Two Types of Business Relationship Customer provider relationship a provider is an AS that connects the customer to the rest of the Internet customer pays the provider for the transit service e.g., Yale is a customer of AT&T and QWEST Peer-to-peer relationship mutually agree to exchange traffic between their respective customers only there is no payment between peers peer provider provider customer provider to customer

Route Selection Policies and Economics Route selection (ranking) policy: the typical route selection policy is to prefer customers over peers/providers to reach a destination, i.e., Customer > pEer/Provider provider customer peer

Export Policies and Economics customer provider peer case 1: routes learned from customer Routes learned from a customer are sent to all other neighbors case 2: routes learned from provider case 3: routes learned from peer provider customer peer peer provider customer Routes learned from a provider are sent only to customers Routes learned from a peer are sent only to customers

Example: Typical Export -> No-Valley Routing IP traffic A advertises path to C, but not P2 A advertises path to C, but not P1 A learns paths to C, P1, P2 A A advertises to C paths to P1 and P2 C Suppose P1 and P2 are providers of A; A is a provider of C

Typical Export Policies Route Patterns Assume a BGP path SABCD to destination AS D. Consider the business relationship between each pair: Three types of business relationships: PC (provider-customer) CP (customer-provider) PP (peer-peer) S A B C D

Typical Export Policies Route Patterns Invariant 1 of valid BGP routes (with labels representing business relationship) P C P C ? P C P C Dest Reasoning: only route learned from customer is sent to provider; thus after a PC, it is always PC to the destination If PC, then PC all the way to destination If CP/PP, all preceding is CP

Typical Export Policies Route Patterns Invariant 2 of valid BGP routes (with labels representing business relationship) CP ······ ? CP /PP Dest Reasoning: routes learned from peer or provider are sent to only customers; thus all relationship before is CP. If PC, then PC all the way to destination If CP/PP, all preceding is CP

Stability of BGP Policy Routing Suppose there is no loop formed by provider-customer relationship in the Internet each AS uses typical route selection policy: C > E/P each AS uses the typical export policies Then policy routing always converges (i.e., is stable).

Case 1: A Link is PC Proof by contradiction. Assume a loop in P-graph. Consider a fixed link. in the loop AS 1 AS 2 AS 3 AS 4 PC PC PC

Case 2: Link is CP/PP AS 1 AS 2 AS 3 AS 4 CP CP CP/PP CP/PP CP/PP

Summary: BGP Policy Routing Issues policy dispute can lead to instability current Internet economy provides a stability framework, but if the framework changes, we may see instability Hierarchical routing can be inefficient AS 4 AS 3 AS 2 AS 1

Outline Admin and recap Network control plane Routing Link weights assignment Routing computation Basic routing protocols (distance vector, link state) Global Internet routing Basic architecture Policy routing analysis Address assignment

IP Addressing: How to Get One? Q: How does an ISP get its block of addresses? A: Local Internet Registry (LIR) or National Internet Registry (NIR) https://www.iana.org/numbers https://www.iana.org/assignments/ipv4-address-space/ipv4-address-space.xhtml Use %whois <IP address> to check who is allocated the given address.

IP addresses: How to Get One? Q: How does a host get an IP address? A: Static configured unix: %/sbin/ifconfig eth0 inet 192.168.0.10 netmask 255.255.255.0 DHCP: Dynamic Host Configuration Protocol (RFC2131): dynamically get address from a DHCP server

DHCP Goal and History Goal: allow host to dynamically obtain its IP address from network server when it joins network History 1984 Reverse ARP (RFC903): obtain IP address, but at link layer, and hence requires a server at each network link 1985 Bootstrap Protocol (BOOTP; RFC951): introduces the concept of a relay agent to forward across networks 1993 DHCP (RFC1531): based on BOOTP but can dynamically allocate and reclaim IP addresses in a pool, as well as delivery of other parameters 1993 Errors in editorials led to immediate reissue as RFC1541 1997 DHCP (RFC2131): add DHCPINFORM

Exercise DHCP wireshark example

DHCP: Dynamic Host Configuration Protocol The often used DORA model (4 messages) host broadcasts “DHCP discover” msg DHCP server responds with “DHCP offer” msg host requests IP address: “DHCP request” msg DHCP server sends address: “DHCP ack” msg

https://www.ietf.org/rfc/rfc2131.txt